Search

In this paper, a new learning-based method is proposed for the early detection of changes of parameters in power converters. It circumvents the pertinent shortcomings of previous model-based methods, such as their need for acquiring switching signals, dependence on the control method, or the need for isolating a part of the system during monitoring and thus interfering with the system performance or its start-up time. Additionally, a downside of the learning-based approaches, which is that their performance depends on the quality of the measurements, is addressed through constructing hybrid models that combine the benefits of both lines of development. Our approaches are evaluated with several types of features based on wavelet decomposition and empirical mode decomposition. Using these, an ANN-based classifier is trained for fault detection. For achieving the final decision on the presence or absence of a fault state, we propose the use of sequential hypothesis testing. This hybrid approach yields a much higher reliability than instantaneous ANN-based classification, while allowing for a statistically sound cross-temporal information integration and providing a controllable error bound for the probability of misclassifications. The proposed method was evaluated on two data-sets recorded from a buck converter and an arm of a modular multilevel converter. The results show that, for both systems, the proposed method is capable of reliably recognizing changes of the system parameters. The potential for practical applications is shown through an implementation on a low-power-consumption microcontroller.

In this paper, we present our recently introduced “assistance on demand (AOD)” concept, which allows the driver to request assistance via speech whenever he or she deems it appropriate. The target scenario we currently investigate is turning left from a subordinate road in dense urban traffic. We first compare our system in a driving simulator study to driving without assistance or with visual assistance. The results show that drivers clearly prefer our speech-based AOD approach. Next, we investigate the differences between drivers’ left-turn behaviour in a driving simulator. The results of this investigation show that there are large inter-individual differences. Based on these results, we present another driving simulator study, where participants can compare manual driving to driving with a default and a personalized AOD system. The results of this second study show that the personalization very notably improves the acceptance of the system. Given the choice between driving with any of the AOD variants and manual driving, 87.5% of the participants preferred driving with the AOD. Finally, we present an evaluation of the AOD system in a prototype vehicle in real urban traffic.

Audio-visual speech recognition (AVSR) can significantly improve performance over audio-only recognition for small or medium vocabularies. However, current AVSR, whether hybrid or end-to-end (E2E), still does not appear to make optimal use of this secondary information stream as the performance is still clearly diminished in noisy conditions for large-vocabulary systems. We, therefore, propose a new fusion architecture—the decision fusion net (DFN). A broad range of time-variant reliability measures are used as an auxiliary input to improve performance. The DFN is used in both hybrid and E2E models. Our experiments on two large-vocabulary datasets, the Lip Reading Sentences 2 and 3 (LRS2 and LRS3) corpora, show highly significant improvements in performance over previous AVSR systems for large-vocabulary datasets. The hybrid model with the proposed DFN integration component even outperforms oracle dynamic stream-weighting, which is considered to be the theoretical upper bound for conventional dynamic stream-weighting approaches. Compared to the hybrid audio-only model, the proposed DFN achieves a relative word-error-rate reduction of 51% on average, while the E2E-DFN model, with its more competitive audio-only baseline system, achieves a relative word error rate reduction of 43%, both showing the efficacy of our proposed fusion architecture.

When a number of speakers are simultaneously active, for example in meetings or noisy public places, the sources of interest need to be separated from interfering speakers and from each other in order to be robustly recognized. Independent component analysis (ICA) has proven a valuable tool for this purpose. However, ICA outputs can still contain strong residual components of the interfering speakers whenever noise or reverberation is high. In such cases, nonlinear postprocessing can be applied to the ICA outputs, for the purpose of reducing remaining interferences. In order to improve robustness to the artefacts and loss of information caused by this process, recognition can be greatly enhanced by considering the processed speech feature vector as a random variable with time-varying uncertainty, rather than as deterministic. The aim of this paper is to show the potential to improve recognition of multiple overlapping speech signals through nonlinear postprocessing together with uncertainty-based decoding techniques.